heddle-objects 0.4.0

// SPDX-License-Identifier: Apache-2.0
//! Tests for packfile operations.

use tempfile::TempDir;

use super::{ObjectType, PackBuilder, PackObjectId, PackReader, pack_index::PackIndex};
use crate::{
    delta::MAX_DELTA_OUTPUT_SIZE,
    object::{ChangeId, ContentHash},
    store::{StoreError, compression::CompressionConfig, pack::pack_container_spec},
};

fn create_test_hash(n: u8) -> ContentHash {
    let bytes: [u8; 32] = [n; 32];
    ContentHash::from_bytes(bytes)
}

fn single_record_pack(
    hash: ContentHash,
    write_record_tail: impl FnOnce(&mut Vec<u8>),
) -> (Vec<u8>, Vec<u8>) {
    let mut pack_data = Vec::new();
    pack_data.extend_from_slice(pack_container_spec().magic);
    pack_data.extend_from_slice(&pack_container_spec().version.to_be_bytes());
    pack_data.extend_from_slice(&1u64.to_be_bytes());

    let entry_offset = u64::try_from(pack_data.len()).expect("test pack offset fits in u64");
    PackObjectId::Hash(hash).encode_tagged(&mut pack_data);
    write_record_tail(&mut pack_data);
    super::append_container_checksum(&mut pack_data);

    let mut index = PackIndex::new();
    index.add(PackObjectId::Hash(hash), entry_offset);
    index.sort();

    (pack_data, index.to_bytes())
}

fn assert_invalid_object_message_contains(error: StoreError, expected: &str) {
    assert!(
        matches!(error, StoreError::InvalidObject(ref message) if message.contains(expected)),
        "expected InvalidObject containing '{expected}', got: {error:?}"
    );
}

#[test]
fn test_pack_container_header_codec_matches_legacy_bytes() {
    let mut legacy = Vec::new();
    legacy.extend_from_slice(b"LMPK");
    legacy.extend_from_slice(&2_u32.to_be_bytes());
    legacy.extend_from_slice(&0_u64.to_be_bytes());
    let checksum = blake3::hash(&legacy);
    legacy.extend_from_slice(checksum.as_bytes());

    let mut encoded = Vec::new();
    super::write_container_header(&mut encoded, pack_container_spec(), 0);
    super::append_container_checksum(&mut encoded);

    assert_eq!(encoded, legacy);
    assert_eq!(
        super::verify_container(&legacy, pack_container_spec()).unwrap(),
        (0, 16, 16)
    );
}

#[test]
fn test_pack_index_header_codec_matches_legacy_bytes() {
    let legacy = b"LMI\0\
        \0\0\0\x02\
        \0\0\0\0\0\0\0\0"
        .to_vec();

    let encoded = PackIndex::new().to_bytes();

    assert_eq!(encoded, legacy);
    assert!(PackIndex::from_bytes(&legacy).unwrap().ids().is_empty());
}

#[test]
fn test_pack_index_roundtrip() {
    let mut index = PackIndex::new();
    index.add(PackObjectId::Hash(create_test_hash(1)), 100);
    index.add(PackObjectId::Hash(create_test_hash(2)), 200);
    index.add(PackObjectId::ChangeId(ChangeId::from_bytes([3; 16])), 300);
    index.sort();

    let bytes = index.to_bytes();
    let restored = PackIndex::from_bytes(&bytes).expect("Failed to deserialize index");

    assert_eq!(
        restored.find(&PackObjectId::Hash(create_test_hash(1))),
        Some(100)
    );
    assert_eq!(
        restored.find(&PackObjectId::Hash(create_test_hash(2))),
        Some(200)
    );
    assert_eq!(
        restored.find(&PackObjectId::ChangeId(ChangeId::from_bytes([3; 16]))),
        Some(300)
    );
    assert_eq!(
        restored.find(&PackObjectId::Hash(create_test_hash(4))),
        None
    );
}

#[test]
fn test_pack_builder_basic() {
    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);

    let hash1 = create_test_hash(1);
    let data1 = b"Hello, World!".to_vec();
    builder.add(hash1, ObjectType::Blob, data1.clone());

    let hash2 = create_test_hash(2);
    let data2 = b"Goodbye, World!".to_vec();
    builder.add(hash2, ObjectType::Blob, data2.clone());

    let (pack_data, index_data, stats) = builder.build().expect("Failed to build pack");

    assert!(!pack_data.is_empty());
    assert!(!index_data.is_empty());
    assert_eq!(stats.object_count, 2);
    assert!(stats.compression_ratio > 0.0 && stats.compression_ratio <= 1.0);
}

#[test]
fn test_pack_reader() {
    let temp_dir = TempDir::new().expect("Failed to create temp dir");
    let pack_path = temp_dir.path().join("test.pack");
    let index_path = temp_dir.path().join("test.idx");

    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);

    let hash1 = create_test_hash(1);
    let data1 = b"Test data 1".repeat(100);
    builder.add(hash1, ObjectType::Blob, data1.clone());

    let (pack_data, index_data, _) = builder.build().expect("Failed to build pack");

    std::fs::write(&pack_path, &pack_data).expect("Failed to write pack file");
    std::fs::write(&index_path, &index_data).expect("Failed to write index file");

    let reader = PackReader::open(&pack_path, &index_path).expect("Failed to open pack");
    let (obj_type, retrieved) = reader
        .get_hashed_object(&hash1)
        .expect("Failed to get object")
        .expect("Object not found");

    assert_eq!(obj_type, ObjectType::Blob);
    assert_eq!(retrieved, data1);
}

#[test]
fn test_delta_compression() {
    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);

    let base_hash = create_test_hash(1);
    let base_data = b"This is the base content. ".repeat(100).to_vec();
    builder.add(base_hash, ObjectType::Blob, base_data.clone());

    let target_hash = create_test_hash(2);
    let target_data = b"This is modified content. ".repeat(100).to_vec();
    builder.add(target_hash, ObjectType::Blob, target_data.clone());

    let (_pack_data, _index_data, stats) = builder.build().expect("Failed to build pack");

    assert!(stats.delta_count > 0);
    assert!(stats.compression_ratio < 1.0);
}

#[test]
fn test_pack_reader_rejects_compressed_size_that_overflows_record_end() {
    let hash = create_test_hash(42);
    let (pack_data, index_data) = single_record_pack(hash, |record| {
        super::varint::encode_type_and_size(ObjectType::Blob, 1, record);
        super::varint::encode_varint(u64::MAX, record);
    });
    let reader = PackReader::from_bytes(pack_data, index_data).expect("container is well-formed");

    let error = reader
        .get_hashed_object(&hash)
        .expect_err("oversized compressed_size must fail before slicing");
    assert!(
        matches!(
            error,
            StoreError::InvalidObject(ref message)
                if message.contains("overflows") || message.contains("platform limits")
        ),
        "expected overflow/platform-limit error, got: {error:?}",
    );

    let bytes_error = reader
        .get_hashed_object_bytes(&hash)
        .expect_err("zero-copy path must reject oversized compressed_size too");
    assert!(
        matches!(
            bytes_error,
            StoreError::InvalidObject(ref message)
                if message.contains("overflows") || message.contains("platform limits")
        ),
        "expected overflow/platform-limit error, got: {bytes_error:?}",
    );
}

#[test]
fn test_pack_reader_rejects_uncompressed_size_above_pack_object_limit() {
    let hash = create_test_hash(48);
    let (pack_data, index_data) = single_record_pack(hash, |record| {
        super::varint::encode_type_and_size(ObjectType::Blob, u64::from(u32::MAX) + 1, record);
        super::varint::encode_varint(1, record);
        record.push(0);
    });
    let reader = PackReader::from_bytes(pack_data, index_data).expect("container is well-formed");

    let error = reader
        .get_hashed_object(&hash)
        .expect_err("absurd uncompressed_size must fail before allocation");
    assert_invalid_object_message_contains(error, "Pack object output size");

    let bytes_error = reader
        .get_hashed_object_bytes(&hash)
        .expect_err("zero-copy path must reject absurd uncompressed_size too");
    assert_invalid_object_message_contains(bytes_error, "Pack object output size");
}

#[test]
fn test_pack_reader_rejects_truncated_compressed_size_varint() {
    let hash = create_test_hash(43);
    let (pack_data, index_data) = single_record_pack(hash, |record| {
        super::varint::encode_type_and_size(ObjectType::Blob, 4, record);
        record.push(0x80);
    });
    let reader = PackReader::from_bytes(pack_data, index_data).expect("container is well-formed");

    let error = reader
        .get_hashed_object(&hash)
        .expect_err("truncated compressed_size must not read into checksum bytes");
    assert_invalid_object_message_contains(error, "Truncated compressed_size varint");

    let bytes_error = reader
        .get_hashed_object_bytes(&hash)
        .expect_err("zero-copy path must reject truncated compressed_size too");
    assert_invalid_object_message_contains(bytes_error, "Truncated compressed_size varint");
}

#[test]
fn test_pack_reader_rejects_compressed_size_past_content_end() {
    let hash = create_test_hash(44);
    let (pack_data, index_data) = single_record_pack(hash, |record| {
        super::varint::encode_type_and_size(ObjectType::Blob, 10, record);
        super::varint::encode_varint(10, record);
        record.extend_from_slice(b"abc");
    });
    let reader = PackReader::from_bytes(pack_data, index_data).expect("container is well-formed");

    let error = reader
        .get_hashed_object(&hash)
        .expect_err("record payload shorter than compressed_size must fail");
    assert_invalid_object_message_contains(error, "Entry data out of bounds");

    let bytes_error = reader
        .get_hashed_object_bytes(&hash)
        .expect_err("zero-copy path must reject payload shorter than compressed_size too");
    assert_invalid_object_message_contains(bytes_error, "Entry data out of bounds");
}

#[test]
fn test_pack_reader_decodes_well_formed_manual_record() {
    let hash = create_test_hash(45);
    let payload = b"manual-pack-record".to_vec();
    let (pack_data, index_data) = single_record_pack(hash, |record| {
        super::varint::encode_type_and_size(ObjectType::Blob, payload.len() as u64, record);
        super::varint::encode_varint(payload.len() as u64, record);
        record.extend_from_slice(&payload);
    });
    let reader = PackReader::from_bytes(pack_data, index_data).expect("container is well-formed");

    let (obj_type, data) = reader
        .get_hashed_object(&hash)
        .expect("well-formed record should decode")
        .expect("record should exist");
    assert_eq!(obj_type, ObjectType::Blob);
    assert_eq!(data, payload);

    let (bytes_type, bytes) = reader
        .get_hashed_object_bytes(&hash)
        .expect("well-formed zero-copy record should decode")
        .expect("record should exist");
    assert_eq!(bytes_type, ObjectType::Blob);
    assert_eq!(bytes.as_ref(), payload.as_slice());
}

#[test]
fn test_pack_reader_rejects_delta_output_above_limit() {
    let temp_dir = TempDir::new().expect("Failed to create temp dir");
    let pack_path = temp_dir.path().join("test.pack");
    let index_path = temp_dir.path().join("test.idx");

    let target_hash = create_test_hash(9);
    let oversized =
        u64::try_from(MAX_DELTA_OUTPUT_SIZE).expect("MAX_DELTA_OUTPUT_SIZE fits in u64") + 1;

    // Build a valid pack with a delta entry whose uncompressed size exceeds the limit
    let mut pack_data = Vec::new();
    pack_data.extend_from_slice(pack_container_spec().magic);
    pack_data.extend_from_slice(&pack_container_spec().version.to_be_bytes());
    pack_data.extend_from_slice(&1u64.to_be_bytes()); // 1 object

    let entry_offset = pack_data.len() as u64;

    PackObjectId::Hash(target_hash).encode_tagged(&mut pack_data);
    super::varint::encode_type_and_size(ObjectType::Delta, oversized, &mut pack_data);
    super::varint::encode_varint(2, &mut pack_data);
    PackObjectId::Hash(create_test_hash(1)).encode_tagged(&mut pack_data);
    pack_data.extend_from_slice(&[0, b'x']);

    let checksum = blake3::hash(&pack_data);
    pack_data.extend_from_slice(checksum.as_bytes());

    let mut index = PackIndex::new();
    index.add(PackObjectId::Hash(target_hash), entry_offset);
    index.sort();

    std::fs::write(&pack_path, &pack_data).expect("Failed to write pack file");
    std::fs::write(&index_path, index.to_bytes()).expect("Failed to write index file");

    let reader = PackReader::open(&pack_path, &index_path).expect("Failed to open pack");
    let error = reader
        .get_hashed_object(&target_hash)
        .expect_err("oversized delta output should fail");

    assert!(
        matches!(error, crate::store::StoreError::InvalidObject(message) if message.contains("Delta output size"))
    );
}

#[cfg(feature = "zstd")]
#[test]
fn test_pack_reader_rejects_compressed_record_claiming_huge_uncompressed_size() {
    let hash = create_test_hash(46);
    let payload = b"small compressed payload".repeat(32);
    let compressed = zstd::encode_all(payload.as_slice(), 3).expect("test zstd compression works");

    let (pack_data, index_data) = single_record_pack(hash, |record| {
        super::varint::encode_type_and_size(ObjectType::Blob, 0xFFFF_FFFF, record);
        super::varint::encode_varint(compressed.len() as u64, record);
        record.extend_from_slice(&compressed);
    });
    let reader = PackReader::from_bytes(pack_data, index_data).expect("container is well-formed");

    let error = reader
        .get_hashed_object(&hash)
        .expect_err("hostile uncompressed_size must be rejected without eager allocation");
    assert_invalid_object_message_contains(error, "Pack object output size");

    let bytes_error = reader
        .get_hashed_object_bytes(&hash)
        .expect_err("zero-copy fallback must reject hostile uncompressed_size too");
    assert_invalid_object_message_contains(bytes_error, "Pack object output size");
}

#[cfg(feature = "zstd")]
#[test]
fn test_pack_decompression_rejects_streaming_output_past_limit() {
    let payload = vec![0xA5; 64 * 1024];
    let compressed = zstd::encode_all(payload.as_slice(), 3).expect("test zstd compression works");
    assert!(
        compressed.len() < 1024,
        "test payload should exercise compressed-small/decompressed-large shape"
    );

    let error = super::shared::decompress_pack_payload_with_limit(&compressed, 0, 32 * 1024)
        .expect_err("streaming output above the cap must fail cleanly");
    assert_invalid_object_message_contains(error, "Pack object output size");
}

#[cfg(feature = "zstd")]
#[test]
fn test_pack_reader_decodes_compressed_object_larger_than_initial_hint() {
    let hash = create_test_hash(47);
    let payload = vec![0x5C; super::shared::PACK_DECOMPRESSION_INITIAL_CAP + 64 * 1024];
    assert!(payload.len() < super::shared::MAX_PACK_OBJECT_OUTPUT_SIZE);
    let compressed = zstd::encode_all(payload.as_slice(), 3).expect("test zstd compression works");
    assert!(
        compressed.len() < payload.len(),
        "manual pack must use the compressed reader path"
    );

    let (pack_data, index_data) = single_record_pack(hash, |record| {
        super::varint::encode_type_and_size(ObjectType::Blob, payload.len() as u64, record);
        super::varint::encode_varint(compressed.len() as u64, record);
        record.extend_from_slice(&compressed);
    });
    let reader = PackReader::from_bytes(pack_data, index_data).expect("container is well-formed");

    let (obj_type, data) = reader
        .get_hashed_object(&hash)
        .expect("large compressed object should decode")
        .expect("record should exist");
    assert_eq!(obj_type, ObjectType::Blob);
    assert_eq!(data, payload);

    let (bytes_type, bytes) = reader
        .get_hashed_object_bytes(&hash)
        .expect("large compressed object should decode through bytes path")
        .expect("record should exist");
    assert_eq!(bytes_type, ObjectType::Blob);
    assert_eq!(bytes.as_ref(), payload.as_slice());
}

#[test]
fn test_pack_index_rejects_impossible_entry_count() {
    let mut bytes = Vec::new();
    bytes.extend_from_slice(super::pack_index::INDEX_MAGIC);
    bytes.extend_from_slice(&super::pack_index::INDEX_VERSION.to_be_bytes());
    bytes.extend_from_slice(&(2_u64).to_be_bytes());
    bytes.extend_from_slice(create_test_hash(1).as_bytes());
    bytes.extend_from_slice(&(123_u64).to_be_bytes());

    let error = match PackIndex::from_bytes(&bytes) {
        Ok(_) => panic!("impossible count should fail"),
        Err(error) => error,
    };
    assert!(
        matches!(error, crate::store::StoreError::InvalidObject(message) if message.contains("count"))
    );
}

#[test]
fn test_pack_reader_rejects_truncated_pack() {
    let temp_dir = TempDir::new().unwrap();
    let pack_path = temp_dir.path().join("test.pack");
    let index_path = temp_dir.path().join("test.idx");

    // Write a pack that's too short (< 32 bytes for checksum)
    std::fs::write(&pack_path, b"short").unwrap();
    std::fs::write(&index_path, b"").unwrap();

    match PackReader::open(&pack_path, &index_path) {
        Err(crate::store::StoreError::InvalidObject(msg)) => {
            assert!(
                msg.contains("too short") || msg.contains("Pack"),
                "expected 'too short' error, got: {msg}"
            );
        }
        Err(e) => panic!("expected InvalidObject, got: {e:?}"),
        Ok(_) => panic!("expected error for truncated pack"),
    }
}

#[test]
fn test_pack_reader_rejects_corrupt_checksum() {
    let temp_dir = TempDir::new().unwrap();
    let pack_path = temp_dir.path().join("test.pack");
    let index_path = temp_dir.path().join("test.idx");

    // Build a valid pack, then corrupt the checksum
    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);
    builder.add(create_test_hash(1), ObjectType::Blob, b"data".to_vec());
    let (mut pack_data, index_data, _) = builder.build().unwrap();

    // Flip a byte in the trailing checksum
    let last = pack_data.len() - 1;
    pack_data[last] ^= 0xFF;

    std::fs::write(&pack_path, &pack_data).unwrap();
    std::fs::write(&index_path, &index_data).unwrap();

    match PackReader::open(&pack_path, &index_path) {
        Err(crate::store::StoreError::InvalidObject(msg)) => {
            assert!(
                msg.contains("checksum"),
                "expected checksum error, got: {msg}"
            );
        }
        Err(e) => panic!("expected InvalidObject, got: {e:?}"),
        Ok(_) => panic!("expected error for corrupt checksum"),
    }
}

#[test]
fn test_pack_index_rejects_bad_magic() {
    let mut bytes = Vec::new();
    bytes.extend_from_slice(b"BAAD"); // wrong magic
    bytes.extend_from_slice(&super::pack_index::INDEX_VERSION.to_be_bytes());
    bytes.extend_from_slice(&0u64.to_be_bytes());

    let err = PackIndex::from_bytes(&bytes).unwrap_err();
    assert!(
        matches!(err, crate::store::StoreError::InvalidObject(ref msg) if msg.contains("magic")),
        "expected magic error, got: {err:?}"
    );
}

#[test]
fn test_pack_index_rejects_bad_version() {
    let mut bytes = Vec::new();
    bytes.extend_from_slice(super::pack_index::INDEX_MAGIC);
    bytes.extend_from_slice(&999u32.to_be_bytes()); // unsupported version
    bytes.extend_from_slice(&0u64.to_be_bytes());

    let err = PackIndex::from_bytes(&bytes).unwrap_err();
    assert!(
        matches!(err, crate::store::StoreError::InvalidObject(ref msg) if msg.contains("version")),
        "expected version error, got: {err:?}"
    );
}

#[test]
fn test_pack_reader_missing_object_returns_none() {
    let temp_dir = TempDir::new().unwrap();
    let pack_path = temp_dir.path().join("test.pack");
    let index_path = temp_dir.path().join("test.idx");

    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);
    builder.add(create_test_hash(1), ObjectType::Blob, b"data".to_vec());
    let (pack_data, index_data, _) = builder.build().unwrap();

    std::fs::write(&pack_path, &pack_data).unwrap();
    std::fs::write(&index_path, &index_data).unwrap();

    let reader = PackReader::open(&pack_path, &index_path).unwrap();
    // Query for a hash that doesn't exist
    let result = reader.get_hashed_object(&create_test_hash(99)).unwrap();
    assert!(result.is_none(), "non-existent hash should return None");
}

/// Stale `.idx` regression: when the index routes a lookup for hash
/// `A` to the on-disk offset of a different record (hash `B`'s
/// position), `get_hashed_object` must reject the result rather than
/// silently returning B's bytes under A's name. The tagged id at
/// each record's start is the cheap authenticator that catches the
/// mismatch — see `PackReader::verify_record_id_matches`.
///
/// Repro: build a pack with two distinct blobs, then synthesize a
/// new index that swaps the two offsets, write it back over the
/// real index, reopen the reader, and assert the lookup errors out
/// with the expected diagnostic phrase.
#[test]
fn stale_index_swapped_offsets_surfaces_as_invalid_object() {
    use crate::store::{StoreError, pack::pack_index::PackIndex};

    let blob_a = b"alpha-payload alpha-payload alpha-payload alpha".to_vec();
    let blob_b = b"bravo-payload bravo-payload bravo-payload bravo".to_vec();
    let hash_a = ContentHash::compute(&blob_a);
    let hash_b = ContentHash::compute(&blob_b);

    let temp_dir = TempDir::new().unwrap();
    let pack_path = temp_dir.path().join("test.pack");
    let index_path = temp_dir.path().join("test.idx");

    let mut builder = PackBuilder::new(CompressionConfig::default());
    builder.add_with_path(hash_a, ObjectType::Blob, blob_a.clone(), None);
    builder.add_with_path(hash_b, ObjectType::Blob, blob_b.clone(), None);
    let (pack_data, index_data, _) = builder.build().unwrap();
    std::fs::write(&pack_path, &pack_data).unwrap();
    std::fs::write(&index_path, &index_data).unwrap();

    // Sanity: untouched index reads cleanly.
    {
        let reader = PackReader::open(&pack_path, &index_path).unwrap();
        let (_, got_a) = reader.get_hashed_object(&hash_a).unwrap().expect("A");
        assert_eq!(got_a, blob_a);
        let (_, got_b) = reader.get_hashed_object(&hash_b).unwrap().expect("B");
        assert_eq!(got_b, blob_b);
    }

    // Synthesize a stale index that swaps the two offsets.
    let original_index = PackIndex::from_bytes(&index_data).unwrap();
    let offset_a = original_index
        .find(&PackObjectId::Hash(hash_a))
        .expect("index has A");
    let offset_b = original_index
        .find(&PackObjectId::Hash(hash_b))
        .expect("index has B");
    assert_ne!(offset_a, offset_b);
    let mut stale = PackIndex::new();
    stale.add(PackObjectId::Hash(hash_a), offset_b); // A → B's offset
    stale.add(PackObjectId::Hash(hash_b), offset_a); // B → A's offset
    stale.sort();
    std::fs::write(&index_path, stale.to_bytes()).unwrap();

    let reader = PackReader::open(&pack_path, &index_path).unwrap();
    let err = reader
        .get_hashed_object(&hash_a)
        .expect_err("stale index must surface as an error, not silent wrong bytes");
    assert!(
        matches!(&err, StoreError::InvalidObject(msg) if msg.contains("stale or corrupt")),
        "expected InvalidObject('… stale or corrupt …'), got: {err:?}",
    );
    // The zero-copy path takes a different fork inside
    // `get_object_bytes`; verify it independently.
    let err_bytes = reader
        .get_hashed_object_bytes(&hash_a)
        .expect_err("zero-copy path must reject too");
    assert!(matches!(err_bytes, StoreError::InvalidObject(_)));
}

/// Helper to write a pack+index to disk and open a reader.
fn build_and_open_pack(
    objects: Vec<(ContentHash, ObjectType, Vec<u8>, Option<String>)>,
) -> PackReader<'static> {
    let temp_dir = TempDir::new().unwrap();
    let pack_path = temp_dir.path().join("test.pack");
    let index_path = temp_dir.path().join("test.idx");

    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);
    for (hash, obj_type, data, path) in objects {
        builder.add_with_path(hash, obj_type, data, path);
    }

    let (pack_data, index_data, _) = builder.build().unwrap();
    std::fs::write(&pack_path, &pack_data).unwrap();
    std::fs::write(&index_path, &index_data).unwrap();

    // Leak temp_dir so files survive for the reader's lifetime
    std::mem::forget(temp_dir);

    PackReader::open(&pack_path, &index_path).unwrap()
}

#[test]
fn test_delta_chain_roundtrip() {
    // Create a chain of 5 objects that each differ slightly from the previous.
    // The builder should chain A→B→C→D→E where each deltas against its predecessor.
    let shared = b"This is shared content that remains constant across all versions. ".repeat(10);
    let mut objects = Vec::new();

    for i in 0..5u8 {
        let mut data = shared.clone();
        data.extend_from_slice(format!("version {i} unique suffix data here").as_bytes());
        let hash = ContentHash::compute(&data);
        objects.push((
            hash,
            ObjectType::Blob,
            data,
            Some("test/file.txt".to_string()),
        ));
    }

    let hashes: Vec<ContentHash> = objects.iter().map(|(h, _, _, _)| *h).collect();
    let originals: Vec<Vec<u8>> = objects.iter().map(|(_, _, d, _)| d.clone()).collect();

    let reader = build_and_open_pack(objects);

    // All objects should round-trip correctly through the chain
    for (i, (hash, expected)) in hashes.iter().zip(originals.iter()).enumerate() {
        let (obj_type, data) = reader
            .get_hashed_object(hash)
            .unwrap_or_else(|e| panic!("Failed to get object {i}: {e}"))
            .unwrap_or_else(|| panic!("Object {i} not found"));
        assert_eq!(obj_type, ObjectType::Blob, "object {i} type mismatch");
        assert_eq!(&data, expected, "object {i} data mismatch");
    }
}

#[test]
fn test_delta_chain_produces_deltas() {
    // Verify that similar objects grouped by path actually produce delta entries
    let shared = b"Shared base content for delta testing. ".repeat(20);
    let mut objects = Vec::new();

    for i in 0..4u8 {
        let mut data = shared.clone();
        data.extend_from_slice(&[i; 32]);
        let hash = ContentHash::compute(&data);
        objects.push((
            hash,
            ObjectType::Blob,
            data,
            Some("src/main.rs".to_string()),
        ));
    }

    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);
    for (hash, obj_type, data, path) in objects {
        builder.add_with_path(hash, obj_type, data, path);
    }

    let (_, _, stats) = builder.build().unwrap();
    assert!(
        stats.delta_count >= 1,
        "expected deltas, got {}",
        stats.delta_count
    );
}

#[test]
fn test_delta_window_pack_bytes_are_stable() {
    let shared = b"Stable pack bytes fixture. ".repeat(16);
    let mut builder = PackBuilder::new(CompressionConfig {
        enabled: false,
        level: 0,
        min_size: usize::MAX,
        max_delta_size: 10_000_000,
    });

    for i in 0..4u8 {
        let mut data = shared.clone();
        data.extend_from_slice(format!("version {i} suffix").as_bytes());
        builder.add_with_path(
            ContentHash::compute(&data),
            ObjectType::Blob,
            data,
            Some("src/stable.rs".to_string()),
        );
    }

    let (pack_data, index_data, stats) = builder.build().unwrap();
    assert!(stats.delta_count > 0);
    assert_eq!(
        blake3::hash(&pack_data).to_string(),
        "4dc15365d17fbdf22ee4c1a8a44ee3414dda6dbdcb93c583536a4629c575e8ca"
    );
    assert_eq!(
        blake3::hash(&index_data).to_string(),
        "60f164bdc2dbb696c68f1c07a0b84e18f678fe915d7d446e43b1966f365b74f9"
    );
}

#[test]
fn test_single_object_no_delta() {
    // A single object should not produce any deltas
    let data = b"solo object content".repeat(50);
    let hash = ContentHash::compute(&data);

    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);
    builder.add(hash, ObjectType::Blob, data.clone());

    let (_, _, stats) = builder.build().unwrap();
    assert_eq!(stats.delta_count, 0);
    assert_eq!(stats.object_count, 1);
}

#[test]
fn test_small_objects_skip_delta() {
    // Objects smaller than MIN_DELTA_SIZE (256 bytes) should not be delta-encoded
    let data1 = b"short object A".to_vec();
    let data2 = b"short object B".to_vec();

    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);
    builder.add_with_path(
        ContentHash::compute(&data1),
        ObjectType::Blob,
        data1,
        Some("tiny.txt".to_string()),
    );
    builder.add_with_path(
        ContentHash::compute(&data2),
        ObjectType::Blob,
        data2,
        Some("tiny.txt".to_string()),
    );

    let (_, _, stats) = builder.build().unwrap();
    assert_eq!(
        stats.delta_count, 0,
        "small objects should not be delta-encoded"
    );
}

#[test]
fn test_chain_resets_on_bad_delta() {
    // When objects are very different, the pack builder may or may not use deltas
    // depending on compression. Either way, all objects must round-trip correctly.
    let data1: Vec<u8> = (0..1024).map(|i| ((i * 131 + 17) % 256) as u8).collect();
    let data2: Vec<u8> = (0..1024).map(|i| ((i * 197 + 53) % 256) as u8).collect();
    let _data3: Vec<u8> = (0..1024).map(|i| ((i * 251 + 89) % 256) as u8).collect();

    // Verify all objects round-trip correctly regardless of delta decisions
    let temp_dir = TempDir::new().unwrap();
    let pack_path = temp_dir.path().join("test.pack");
    let index_path = temp_dir.path().join("test.idx");
    let mut builder2 = PackBuilder::new(CompressionConfig::default());
    builder2.add_with_path(
        ContentHash::compute(&data1),
        ObjectType::Blob,
        data1.clone(),
        Some("file.bin".to_string()),
    );
    builder2.add_with_path(
        ContentHash::compute(&data2),
        ObjectType::Blob,
        data2.clone(),
        Some("file.bin".to_string()),
    );
    let (pd, id, _) = builder2.build().unwrap();
    std::fs::write(&pack_path, &pd).unwrap();
    std::fs::write(&index_path, &id).unwrap();
    let reader = PackReader::open(&pack_path, &index_path).unwrap();
    let (_, got) = reader
        .get_hashed_object(&ContentHash::compute(&data1))
        .unwrap()
        .unwrap();
    assert_eq!(got, data1);
    let (_, got) = reader
        .get_hashed_object(&ContentHash::compute(&data2))
        .unwrap()
        .unwrap();
    assert_eq!(got, data2);
}

#[test]
fn test_different_paths_with_different_content_roundtrip() {
    // Objects with different paths and truly different content should round-trip correctly.
    // The sliding window may or may not produce deltas depending on content similarity —
    // what matters is that all objects are retrievable.
    let base_a = vec![0xAA; 1024]; // completely different byte patterns
    let base_b = vec![0xBB; 1024];

    let hash_a = ContentHash::compute(&base_a);
    let hash_b = ContentHash::compute(&base_b);

    let reader = build_and_open_pack(vec![
        (
            hash_a,
            ObjectType::Blob,
            base_a.clone(),
            Some("a.bin".to_string()),
        ),
        (
            hash_b,
            ObjectType::Blob,
            base_b.clone(),
            Some("b.bin".to_string()),
        ),
    ]);

    let (_, got_a) = reader.get_hashed_object(&hash_a).unwrap().unwrap();
    assert_eq!(got_a, base_a);
    let (_, got_b) = reader.get_hashed_object(&hash_b).unwrap().unwrap();
    assert_eq!(got_b, base_b);
}

#[test]
fn test_objects_without_path_use_size_bucketing() {
    // Objects without path hints should fall back to size-based bucketing
    let shared = b"Shared prefix for size bucketing test with enough content. ".repeat(10);

    let mut objects = Vec::new();
    for i in 0..3u8 {
        let mut data = shared.clone();
        data.extend_from_slice(&[i; 16]);
        objects.push((ContentHash::compute(&data), ObjectType::Blob, data, None));
    }

    let originals: Vec<(ContentHash, Vec<u8>)> =
        objects.iter().map(|(h, _, d, _)| (*h, d.clone())).collect();

    let reader = build_and_open_pack(objects);

    for (hash, expected) in &originals {
        let (_, data) = reader.get_hashed_object(hash).unwrap().unwrap();
        assert_eq!(&data, expected);
    }
}

#[test]
fn test_tree_objects_can_be_delta_encoded() {
    // Tree objects should also support delta encoding
    let shared = b"tree serialization data that is shared ".repeat(15);

    let mut objects = Vec::new();
    for i in 0..3u8 {
        let mut data = shared.clone();
        data.extend_from_slice(format!("tree version {i}").as_bytes());
        objects.push((
            ContentHash::compute(&data),
            ObjectType::Tree,
            data,
            Some("src/".to_string()),
        ));
    }

    let originals: Vec<(ContentHash, Vec<u8>)> =
        objects.iter().map(|(h, _, d, _)| (*h, d.clone())).collect();

    let reader = build_and_open_pack(objects);

    for (hash, expected) in &originals {
        let (obj_type, data) = reader.get_hashed_object(hash).unwrap().unwrap();
        assert_eq!(obj_type, ObjectType::Tree);
        assert_eq!(&data, expected);
    }
}

#[test]
fn test_state_objects_not_delta_encoded() {
    // State objects should never be delta-encoded (guard in group_by_type)
    let data1 = b"state data 1".repeat(50);
    let data2 = b"state data 2".repeat(50);

    let compression = CompressionConfig::default();
    let mut builder = PackBuilder::new(compression);
    builder.add(ContentHash::compute(&data1), ObjectType::State, data1);
    builder.add(ContentHash::compute(&data2), ObjectType::State, data2);

    let (_, _, stats) = builder.build().unwrap();
    assert_eq!(stats.delta_count, 0, "states should never be delta-encoded");
}

#[test]
fn test_empty_bucket_is_noop() {
    // An empty pack should produce valid but minimal output
    let compression = CompressionConfig::default();
    let builder = PackBuilder::new(compression);
    let (pack_data, _, stats) = builder.build().unwrap();

    assert_eq!(stats.object_count, 0);
    assert_eq!(stats.delta_count, 0);
    // Pack data should still have magic + version + count + checksum
    assert!(pack_data.len() >= 16 + 32); // header + checksum
}